An open source machine learning platform for managing the complete ML lifecycle, developed at Databricks, that includes four components supporting experimentation, reproducibility, deployment, and a central model registry.

Use cases for MLflow include:

Generative AI

Improve generative AI quality
Build applications with prompt engineering
Track progress during fine tuning
Package and deploy models
Securely host LLMs at scale with MLflow Deployments

Deep Learning

Native integrations with popular DL frameworks (PyTorch, TensorFlow, Keras)
Simple, low-code performance tracking with autologging
UI for deep learning model analysis and comparison

Traditional Machine Learning

End-to-end MLOps solution for traditional ML, including integrations with scikit-learn, XGBoost, and PySpark
Simple, low-code performance tracking with autologging
UI for model analysis and comparison

Evaluation

Compare different ML models and GenAI application versions
Evaluate different prompts
Compare performance against a baseline to prevent regressions
Simplify and automate performance evaluation

Model Management

Package models for production, including code and dependencies
Catalog, govern, and manage model versions
Orchestrate model rollouts to staging and production
Deploy models for large scale batch and real-time inference